While it is clear that the ability of firms to introduce new products or processes is crucially affected by their choice of innovation partners, the geographical dimension of these partnerships has until fairly recently attracted relatively little attention. Yet, the factors which drive firms to collaborate with far away agents or, by contrast, to predominantly interact locally are still poorly understood. Hence questions such as which factors affect the use by firms of partners at different geographical distances, or whether the use of partners in different geographical locations depends only or mainly on factors internal to the firm or on the regional environment remain unanswered. In this paper we examine which are the factors behind the geographical dimension of interactive learning. We assess to what extent the propensity of firms to establish collaborations at different geographical distances depends on three types of factors. These factors include manager-level, firm-level and regional-level variables. Using data stemming from a survey of 1604 businesses located in five Norwegian city-regions, we model firms' use of partners located within the region, elsewhere in the country, and abroad, respectively. The results indicate that collaboration and interactive learning are affected by variables related to all three levels, but that these mechanisms have radically different effects on local compared to international collaboration. At the level of the manager, trust is an important predictor of regional and national collaboration, but has no significant effect on the formation of international partnerships, which are fundamentally associated with factors such as education and the open-mindedness of managers. At the firm level, size tends to have a positive effect, but more so for international than regional collaboration. Foreign ownership also has a positive effect on international collaboration, but negative on regional collaboration. At the regional level, R&D expenditure tends to increase collaboration between regional actors, but reduces the likelihood of engagement with international partners. Education, by contrast, has the opposite effect: it encourages international collaboration at the expense of local links. The results highlight the need to balance policies for boosting regional social capital and R&D with investments in education and encouragement of open-mindedness in order to ensure sufficient construction of global pipelines and avoid over-reliance on local buzz.